Inference interractions

Note

Please, click on this link to get back to the Home Page.

Here I will compare results of packages HMSC, gllvm, ecoCopula and EMtree.

HMSC

First, let’s try HMSC. We don’t have informations about sampling design, let’s assume that each site have been visited once.

Now, we will assume that there is no interactions between covariables.

Then we set up the MCMC chains.

And run the model.

Then we diagnostic the model.

Not so great for chain convergence…

Let’s see the correlation with the environnemtal variables.

There is no correlation with the species and the bare sediment and reflexion.

So there is no inter-specific interactions.

ecoCopula

It’s making a small graph, some species do not seems to interact.

EMTree

We’ll start with the same model


 Initialization...
 Adjusting a PLN model with full covariance model
 Post-treatments...
 DONE!

Likelihoods: 17.61764 , 17.6573 , 17.65734 , 


Convergence took 0.27 secs  and  3  iterations.
Likelihood difference = 4.241987e-05 
Betas difference = 8.172318e-08
List of 5
 $ edges_prob  : num [1:12, 1:12] 0 0.2777 0.1111 0.1336 0.0908 ...
 $ edges_weight: num [1:12, 1:12] 0 0.00761 0.00756 0.00757 0.00755 ...
 $ logpY       : num [1:3] 17.6 17.7 17.7
 $ maxIter     : num 3
 $ timeEM      : 'difftime' num 0.26788592338562
  ..- attr(*, "units")= chr "secs"
$G


$graph_data
# A tbl_graph: 12 nodes and 20 edges
#
# An undirected simple graph with 2 components
#
# Node Data: 12 x 8 (active)
    btw bool_btw bool_deg   deg title name     label      finalcolor
  <dbl> <lgl>    <lgl>    <dbl> <chr> <chr>    <chr>      <lgl>     
1     8 FALSE    TRUE         4 ""    Alopacce "Alopacce" FALSE     
2     0 FALSE    TRUE         4 ""    Alopcune "Alopcune" FALSE     
3     0 FALSE    TRUE         2 ""    Alopfabr "Alopfabr" FALSE     
4     0 FALSE    TRUE         2 ""    Arctlute "Arctlute" FALSE     
5     0 FALSE    FALSE        0 ""    Arctperi ""         FALSE     
6     0 FALSE    TRUE         2 ""    Auloalbi "Auloalbi" FALSE     
# … with 6 more rows
#
# Edge Data: 20 x 6
   from    to weight btw.weights neibs title
  <int> <int>  <dbl>       <dbl> <lgl> <chr>
1     1     2  0.278        1.53 FALSE ""   
2     1     8  0.369        1.31 FALSE ""   
3     1     9  0.362        1.32 TRUE  ""   
# … with 17 more rows

2020-01-17